POI quality and usage – any corporates contributing?

I’ve been contributing to POI data for probably two years now, but as a map user I honestly find it quite hit and miss and lacking.

Why is that, and what could be done to improve it?

It’s only updated if there’s someone like me in the local area, but even me have only really been able to “complete” two shopping streets in my city in Oslo, Norway.

Tellingly, Overture Maps chose to use OSM for the street data, but not the POI data.

A lot of people pointed out that the POI data of Overture had mediocre quality. While that is true, they do at least have coverage in places OSM do not. If you know the name of the place you can at least find it

I’m not sure how this can become better, I’d think that a prerequisite would be that corporate users like Uber, Grab, DiDi, TomTom start contributing.

But that may not happen as they are not allowed to combine datasets. Hence you have a chicken and egg problem: The dataset is not usable until it is complete…

The street data fundementally have the same problem, but that is easier to edit remotely and may be more stable in contrast to POIs, such that it overcame this hurdle

So I’m starting to question if it’s really worth it to continue update POIs in OSM and my time would be better spent on other efforts?

Are any corporate users using and contributing to the POI dataset today?


I remember a mention in the WeeklyOSM from over a year ago that mentioned an initiative by a group of 500 French businesses to map and maintain their POIs on OSM. There is also OSM user Borishag who is employed by some Dutch store chains to maintain their presence on OSM.

Many business owners (or marketing agencies that they hire) also map their business with new accounts that just map a single POI and are then abandoned. That’s how we get SEO spam. But remove or ignore the SEO and suddenly you have thousands of valuable POI contributions from business owners and operators.

Obviously none of this comes close to what Google currently has, but I believe we can leverage POIs from Overture to change that gradually.

One thing I’ve found is that adding POIs remotely becomes much easier once there’s good address coverage in a place.

I was able to add most of the businesses in a town in Massachusetts in a few hours from my couch using addresses from their websites (and a combination of local knowledge and streetside to know what businesses to look for), because Massachusetts has pretty complete address coverage. Then next time I visited, it was a quick drive around town to sanity check the website data and find any changes that had happened. I tried do that in another town I’m familiar with in Minnesota where address coverage is a lot more sparse, and it was much more difficult.

The other thing to keep in mind is that “The dataset is not usable until it is complete” is relative. In my experience OSM has pretty great POI coverage of public places like schools and parks — which is very useable data already. Keeping that in mind, it may make sense to focus on another particular class of POIs over a wider area, and get that class closer to “complete” and useable, rather than focusing on adding every business at once.

1 Like

I find it valuable to contribute POI data where it’s low effort or especially useful.

Low effort: when I’m waiting for the bus I open Every Door and update nearby POIs. Same thing at the airport. When I’m walking to a shop I check for open notes and sometimes take small detours to go to one. This doesn’t really take time away from other activities.

Especially useful: when I’m out on a hike and there’s only a POI every couple of miles I make sure it’s on the map. (I take a geotagged photo and when I have a spare minute later, I add it) Popular hiking apps show these POIs, so this reaches far more people than making sure every company sharing an office building is mapped.

I also enjoy using OSM data to answer questions like “where is the nearest X” (e.g. water_park) so personally I find rare POIs more valuable and interesting than common ones.


I had a look at the corporate contribution pages, and TomTom state they do POI edits if they get map feedback about it, while Expedia say they wish to contribute to the hotel POIs

Mapbox seems to have their own proprietary POI database with 170M points described here, and viewable here

In addition there’s Overture with 60M POIs? Why they went with the Meta and Facebook dataset is discussed here, one aspect being that you can’t know how recent POIs are

Hopefully the latter can get more adoption via Organic Maps that seems to be willing to implement check_date, so one small step towards making the places dataset as usable as the road network…

(Yes, I know # of POIs is not a quality metric, just for reference)

Rideshare providers

Map providers


Note https://osmfoundation.org/wiki/Licensing_Working_Group/Minutes/2023-08-14#Ticket#2023081110000064_—_First_party_websites_as_sources

So it seems we can use data straight from shop websites, including from chain websites.

I see big potential here if we can use that.

of unusuably low quality, at least in areas I could investigate

everyone else who tried looking at data quality reported that it is terrible

That is definitely good news that All The Places can be used as a source for OSM!

I see they have 4.7M POI points, how many do OSM have for common POI categories (not amenities like toilets etc., where OSM is far better than any other source?)

Yes, number of POI points is not a quality indicator in itself, and yes Overture is also quite bad quality in my own neighborhood because I have done a good job of cleaning up the stale OSM data :wink:

But in places where there are less OSM contributors (e.g., “most” of the world), Overture will likely be better as OSM is often an almost blank slate… And given the usage of corporates, it will likely improve over time also in the places OSM has the upper hand today?

That is why I’m asking this question – are there anything that can be done to change that such that OSM POI data is not just a curiosity for people liking open data apps and niche use cases like hiking, in contrast to the road dataset which is widely used :slight_smile:

1 Like

(a) that is not true (at least not all data) and (b) the LWG have not said that,.as I said previously here.

A flippant answer might be “rouse yourself from behind your computer keyboard and start surveying”?

More seriously - there will always be different sources of different quality data available. We know that if you do random scraping you end up with data of questionable quality. See this earlier thread for chapter and verse on that.

To see what corporates have available to contribute, see for example here. Selecting 75% confidence and looking at the two data releases there shows that it is getting better (they’ve noted that “Max Car Care Limited” does not operate from the north transept of York Minster, but still the majority of the POIs shown there are “obviously wrong” in some way, even in the latest data release.

If there genuinely is no local surveyed data, then consumers have no choice but to go with whatever other people might have guessed at, but much of that will be wrong - it’s just quite literally “better than nothing”.

Hence my flippant answer above :slight_smile:

This was being discussed on Discord yesterday & as I asked there, does this also suggest that copying a shopping centre directory / map such as https://robinatowncentre.qicre.com/Directory or https://robinatowncentre.qicre.com/Centre-Information/Centre-Map#/ would also be acceptable?

First off: IANAL.

The LWG position statement relies on the principle that raw facts don’t enjoy copyright protection. To the extent that database rights protect raw facts, they describe a common and relevant situation in which database rights don’t apply either. But a map isn’t just a rote list of raw facts; the facts are selected and presented in a certain manner similar to how OSM presents its data. So if you copy the information into OSM such that that part of OSM looks much more like the shopping center map, you’d be taking on more risk.

On the other hand, I happen to own a map that lists all the addresses and emergency phone numbers for the highway patrol of that region. If I publish a spreadsheet of these numbers, that’s a very different presentation of the same raw facts. Hope that helps to understand the distinction, even if it isn’t very precise and definitive.